PROJECT ABSTRACT Epilepsy surgery involving the anteromedial temporal lobe (ATL) often results in deficits of language or memory. The best clinical predictor of postsurgical cognitive deficits is the extent to which tissues around the epileptogenic focus were spared during resection. We recently showed that resting-state functional connectivity (rs-FC) can predict normative variance in higher-order cognition; rs-FC has subsequently been shown to outperform lesion-based analyses for modeling stroke-related deficits of higher-order cognitions including memory. The proposed retrospective analyses of existing datasets will evaluate the prognostic value of rs-FC for modeling presurgical deficits of language and memory among patients with refractory epilepsy. Aim 1 will replicate our methodology for relating rs-FC to cognitive ability using an independent normative sample of healthy control subjects acquired at Emory University. Aim 2 will evaluate the clinical utility of our method by modeling the influence of clinical symptoms (such as seizure frequency and laterality of epileptogenic focus) upon the relationship between rs-FC and presurgical memory and/or language deficits within a sample of patients with refractory epilepsy. Aim 3 will use machine learning approaches to evaluate rs-FC's ability to explain presurgical cognitive deficits relative to other MRI modalities (such as anatomic volume, cortical thickness, white matter integrity, and task-based fMRI activity). The proposed model-based and hypothesis- driven research plan would critically evaluate resting-state fMRI's prognostic capability to explain epilepsy- related cognitive deficits, and thus establish the necessary methodological groundwork for future prospective analyses using rs-FC to predict postsurgical cognitive outcomes.